JAL-1260 refactored GenBank (and EMBL) flat file parser

[jalview.git] / src / jalview / io / GenBankFile.java

diff --git a/src/jalview/io/GenBankFile.java b/src/jalview/io/GenBankFile.java
new file mode 100644 (file)
index 0000000..7988764
--- /dev/null
+++ b/src/jalview/io/GenBankFile.java
@@ -0,0 +1,207 @@
+package jalview.io;
+
+import java.io.IOException;
+
+import jalview.bin.Cache;
+
+/**
+ * A class that provides selective parsing of the GenBank flatfile format.
+ * <p>
+ * The initial implementation is limited to extracting fields used by Jalview
+ * after fetching an EMBL or EMBLCDS entry:
+ * 
+ * <pre>
+ * accession, version, sequence, xref
+ * and (for CDS feature) location, protein_id, product, codon_start, translation
+ * </pre>
+ * 
+ * @author gmcarstairs
+ * @see https://www.ncbi.nlm.nih.gov/Sitemap/samplerecord.html
+ */
+public class GenBankFile extends FlatFile
+{
+  private static final String DEFINITION = "DEFINITION";
+
+  /**
+   * Constructor given a data source and the id of the source database
+   * 
+   * @param fp
+   * @param sourceId
+   * @throws IOException
+   */
+  public GenBankFile(FileParse fp, String sourceId) throws IOException
+  {
+    super(fp, sourceId);
+  }
+
+  /**
+   * Parses the flatfile, and if successful, saves as an annotated sequence
+   * which may be retrieved by calling {@code getSequence()}
+   * 
+   * @throws IOException
+   * @see https://www.ncbi.nlm.nih.gov/Sitemap/samplerecord.html
+   */
+  @Override
+  public void parse() throws IOException
+  {
+    String line = nextLine();
+    while (line != null)
+    {
+      if (line.startsWith(DEFINITION))
+      {
+        line = parseDefinition(line);
+      }
+      else if (line.startsWith("ACCESSION"))
+      {
+        this.accession = line.split(WHITESPACE)[1];
+        line = nextLine();
+      }
+      else if (line.startsWith("VERSION"))
+      {
+        line = parseVersion(line);
+      }
+      else if (line.startsWith("ORIGIN"))
+      {
+        line = parseSequence();
+      }
+      else if (line.startsWith("FEATURES"))
+      {
+        line = nextLine();
+        while (line.startsWith(" "))
+        {
+          line = parseFeature(line);
+        }
+      }
+      else
+      {
+        line = nextLine();
+      }
+    }
+    buildSequence();
+  }
+
+  /**
+   * Extracts and saves the primary accession and version (SV value) from an ID
+   * line, or null if not found. Returns the next line after the one processed.
+   * 
+   * @param line
+   * @throws IOException
+   */
+  String parseLocus(String line) throws IOException
+  {
+    String[] tokens = line.substring(2).split(";");
+
+    /*
+     * first is primary accession
+     */
+    String token = tokens[0].trim();
+    if (!token.isEmpty())
+    {
+      this.accession = token;
+    }
+
+    /*
+     * second token is 'SV versionNo'
+     */
+    if (tokens.length > 1)
+    {
+      token = tokens[1].trim();
+      if (token.startsWith("SV"))
+      {
+        String[] bits = token.trim().split(WHITESPACE);
+        this.version = bits[bits.length - 1];
+      }
+    }
+
+    /*
+     * seventh token is 'length BP'
+     */
+    if (tokens.length > 6)
+    {
+      token = tokens[6].trim();
+      String[] bits = token.trim().split(WHITESPACE);
+      try
+      {
+        this.length = Integer.valueOf(bits[0]);
+      } catch (NumberFormatException e)
+      {
+        Cache.log.error("bad length read in flatfile, line: " + line);
+      }
+    }
+
+    return nextLine();
+  }
+
+  /**
+   * Reads sequence description from DEFINITION lines. Any trailing period is
+   * discarded. Returns the next line after the definition line(s).
+   * 
+   * @param line
+   * @return
+   * @throws IOException
+   */
+  String parseDefinition(String line) throws IOException
+  {
+    String desc = line.substring(DEFINITION.length()).trim();
+    if (desc.endsWith("."))
+    {
+      desc = desc.substring(0, desc.length() - 1);
+    }
+
+    /*
+     * pass over any additional DE lines
+     */
+    while ((line = nextLine()) != null)
+    {
+      if (line.startsWith(" "))
+      {
+        // definition continuation line
+        desc += line.trim();
+      }
+      else
+      {
+        break;
+      }
+    }
+    this.description = desc;
+
+    return line;
+  }
+
+  /**
+   * Parses the VERSION line e.g.
+   * 
+   * <pre>
+   * VERSION     X81322.1
+   * </pre>
+   * 
+   * and returns the next line
+   * 
+   * @param line
+   * @throws IOException
+   */
+  String parseVersion(String line) throws IOException
+  {
+    /*
+     * extract version part of <accession>.<version>
+     * https://www.ncbi.nlm.nih.gov/Sitemap/samplerecord.html#VersionB
+     */
+    String[] tokens = line.split(WHITESPACE);
+    if (tokens.length > 1)
+    {
+      tokens = tokens[1].split("\\.");
+      if (tokens.length > 1)
+      {
+        this.version = tokens[1];
+      }
+    }
+
+    return nextLine();
+  }
+
+  @Override
+  protected boolean isFeatureContinuationLine(String line)
+  {
+    return line.startsWith("      "); // 6 spaces
+  }
+}